Working definition
Critical thinking on AI is the habit of confronting model shortcuts so quality rises — you stay the editor of truth for your stack.
How it works
Models optimise for fluency; you optimise for decisions you must own. Orchestration without challenge ships trash faster.
Example
You paste a client offer into the model and ship the first fluent draft because it sounds decisive — then discover the pricing logic contradicts the Battle you signed with yourself. Fluency hid the missing counter-questions. Three challenges before accept: what evidence? what alternative? what fails for this avatar? That is B2 miss — stay Real Boss of the stack; do not outsource the signature to the first confident reply.
Signature
Refuse the first confident reply when stakes are high — counter, refine, then wire.
Your move this week
Take one AI output this week and write three counter-questions before you accept it.
Scientific grounding
Cite
- Soll, J. B., Milkman, K. L., & Payne, J. W. (2015). A user’s guide to debiasing. In G. Keren & G. Wu…. The Wiley Blackwell handbook of judgment and decision making (pp. 924–951). Wiley.
- Koriat, A., Ackerman, R., & Adiv, S. (2014). The effects of goal-driven and data-driven regulation on metacognitive monitoring during learning. Journal of Experimental Psychology: General, 143(1), 386–403.
Challenge the model is metacognitive monitoring + debiasing — fluency is not truth; you stay the editor.
Adjacent
One Person Business · Quality of Questions · Be Your Real Boss · Coaching (MoC)